Method and system for real-time large-scale scene three-dimensional scanning modeling

Abstract

The invention relates to a method and a system for real-time large-scale scene three-dimensional scanning modeling. The method comprises the following steps: an RGB-D video stream is acquired; global camera pose optimization is carried out on the above video stream to acquire a global camera prediction position; and according to the global camera prediction position, globally-optimized three-dimensional scanning modeling is carried out. Real-time and terminal-to-terminal modeling can be realized. Local camera tracking does not need to rely on, but the globally-optimized camera positioning is kept constantly. A parallel optimization modeling system combining sparse feature points, dense geometry, and color information is brought forward and developed, the globally-optimized camera position can be predicted in real time, camera repositioning is supported, a globally-consistent optimal three-dimensional model can be updated in real time, the system of the invention is applied to large-scale indoor scanning, and a high-quality three-dimensional model can be acquired.

Description

technical field [0001] The present invention relates to the field of three-dimensional modeling, in particular, the present invention relates to a method and system for real-time large-scale scene three-dimensional scanning modeling. Background technique [0002] Real-time high-quality 3D scanning of large-scale scenes is an important application in the field of virtual augmented reality and robotics. The popularity of RGB-D cameras has brought about a huge revolution in 3D scanning. 3D printing, virtual and augmented reality, games, and robotics fields have important requirements for large-scale real-time 3D scanning: robots or ordinary users can instantly obtain synthetic 3D models through real-time scanning, which can be applied to robot navigation and integrate the physical world Introduce a virtual environment, or provide real-time information feedback to users. [0003] However, although 3D reconstruction systems have been developed for a long time, there is still no...

AI Technical Summary

Problems solved by technology

Real-time correction is the key to ensuring the consistency of the global model, which poses a huge challenge

[0007] 4. Stable camera tracking: In addition to the increased error per frame, camera tracking may fail in areas where features are not obvious

Updating the model while updating the camera position is a huge challenge

[0009] 6. Real-time: Fast real-time is an indispensable element in many virtual / augmented reality and robotics applications, which poses a huge challenge

More recent online modeling methods usually have the following problems: (1) still need several minutes of correction, thus can not achieve real-time

(2) The frame-to-frame or frame-to-model camera tracking method is very fragile and often leads to irreparable errors

(3) Only supports the representation method of unstructured point cloud, which seriously affects the quality and practicability of scanning

Images